CITS4009 Project 1

Student Name: Xiao ZHANG

Student Number: 23363869

URL of Video: YouTube Video

1 Introduction

The data set analyzed can be obtained from the Kaggle platform. A collection of YouTube giants, this dataset offers a perfect avenue to analyze and gain valuable insights from the luminaries of the platform. With comprehensive details on top creators’ subscriber counts, video views, upload frequency, country of origin, earnings, and more.Here is the original data source global youtube statisctic.

2 Data loading, overview and set up

Load needed libraries

library(ggplot2)
library(ggthemes)
library(dplyr)
library(shiny)
library(maps)
library(hexbin)
library(countrycode)
library(rworldmap)
library(leaflet)
library(plotly)
library(janitor)

Set up a consistent plot format

plot_theme <- theme_few() + 
              theme(plot.title = element_text(color = "darkred",hjust = 0.5)) +
              theme(strip.text.x = element_text(size = 14, colour = "#202020"))

And then read the data through read.csv function.

youtube_data <- read.csv('youtube_UTF_8.csv',encoding = "UTF-8")

Look at the structure of the data.

str(youtube_data)

## 'data.frame':    995 obs. of  28 variables:
##  $ rank                                   : int  1 2 3 4 5 6 7 8 9 10 ...
##  $ Youtuber                               : chr  "T-Series" "YouTube Movies" "MrBeast" "Cocomelon - Nursery Rhymes" ...
##  $ subscribers                            : int  245000000 170000000 166000000 162000000 159000000 119000000 112000000 111000000 106000000 98900000 ...
##  $ video.views                            : num  2.28e+11 0.00 2.84e+10 1.64e+11 1.48e+11 ...
##  $ category                               : chr  "Music" "Film & Animation" "Entertainment" "Education" ...
##  $ Title                                  : chr  "T-Series" "youtubemovies" "MrBeast" "Cocomelon - Nursery Rhymes" ...
##  $ uploads                                : int  20082 1 741 966 116536 0 1111 4716 493 574 ...
##  $ Country                                : chr  "India" "United States" "United States" "United States" ...
##  $ Abbreviation                           : chr  "IN" "US" "US" "US" ...
##  $ channel_type                           : chr  "Music" "Games" "Entertainment" "Education" ...
##  $ video_views_rank                       : int  1 4055159 48 2 3 4057944 5 44 630 8 ...
##  $ country_rank                           : num  1 7670 1 2 2 NaN 3 1 5 5 ...
##  $ channel_type_rank                      : num  1 7423 1 1 2 ...
##  $ video_views_for_the_last_30_days       : num  2.26e+09 1.20e+01 1.35e+09 1.98e+09 1.82e+09 ...
##  $ lowest_monthly_earnings                : num  564600 0 337000 493800 455900 ...
##  $ highest_monthly_earnings               : num  9.0e+06 5.0e-02 5.4e+06 7.9e+06 7.3e+06 ...
##  $ lowest_yearly_earnings                 : num  6.8e+06 4.0e-02 4.0e+06 5.9e+06 5.5e+06 ...
##  $ highest_yearly_earnings                : num  1.08e+08 5.80e-01 6.47e+07 9.48e+07 8.75e+07 ...
##  $ subscribers_for_last_30_days           : num  2e+06 NaN 8e+06 1e+06 1e+06 NaN NaN NaN 1e+05 6e+05 ...
##  $ created_year                           : num  2006 2006 2012 2006 2006 ...
##  $ created_month                          : chr  "Mar" "Mar" "Feb" "Sep" ...
##  $ created_date                           : num  13 5 20 1 20 24 12 29 14 23 ...
##  $ Gross.tertiary.education.enrollment....: num  28.1 88.2 88.2 88.2 28.1 NaN 88.2 63.2 81.9 88.2 ...
##  $ Population                             : num  1.37e+09 3.28e+08 3.28e+08 3.28e+08 1.37e+09 ...
##  $ Unemployment.rate                      : num  5.36 14.7 14.7 14.7 5.36 NaN 14.7 2.29 4.59 14.7 ...
##  $ Urban_population                       : num  4.71e+08 2.71e+08 2.71e+08 2.71e+08 4.71e+08 ...
##  $ Latitude                               : num  20.6 37.1 37.1 37.1 20.6 ...
##  $ Longitude                              : num  79 -95.7 -95.7 -95.7 79 ...

The dataset is a data frame with 995 observations and 28 variables. These 28 variables cover various types of information, including factor, integer, and numeric variables. Among these 28 variables, there are 7 factor variables, where “Youtuber” and “category” are two factors with a large number of levels. This can mean that there are many different YouTubers and categories of videos. In addition, the dataset also includes 4 integer variables that may involve count-related information such as “rank” and “subscribers”. There are also 17 numeric variables, some of which may be floating point, covering various statistical and quantitative data, such as “video.views”, “lowest_monthly_earnings” and “Population”.

Use summary function to obtain descriptive statistics

summary(youtube_data)

##       rank         Youtuber          subscribers         video.views       
##  Min.   :  1.0   Length:995         Min.   : 12300000   Min.   :0.000e+00  
##  1st Qu.:249.5   Class :character   1st Qu.: 14500000   1st Qu.:4.288e+09  
##  Median :498.0   Mode  :character   Median : 17700000   Median :7.761e+09  
##  Mean   :498.0                      Mean   : 22982412   Mean   :1.104e+10  
##  3rd Qu.:746.5                      3rd Qu.: 24600000   3rd Qu.:1.355e+10  
##  Max.   :995.0                      Max.   :245000000   Max.   :2.280e+11  
##                                                                            
##    category            Title              uploads           Country         
##  Length:995         Length:995         Min.   :     0.0   Length:995        
##  Class :character   Class :character   1st Qu.:   194.5   Class :character  
##  Mode  :character   Mode  :character   Median :   729.0   Mode  :character  
##                                        Mean   :  9187.1                     
##                                        3rd Qu.:  2667.5                     
##                                        Max.   :301308.0                     
##                                                                             
##  Abbreviation       channel_type       video_views_rank   country_rank   
##  Length:995         Length:995         Min.   :      1   Min.   :   1.0  
##  Class :character   Class :character   1st Qu.:    323   1st Qu.:  11.0  
##  Mode  :character   Mode  :character   Median :    916   Median :  51.0  
##                                        Mean   : 554249   Mean   : 386.1  
##                                        3rd Qu.:   3584   3rd Qu.: 123.0  
##                                        Max.   :4057944   Max.   :7741.0  
##                                        NA's   :1         NA's   :116     
##  channel_type_rank video_views_for_the_last_30_days lowest_monthly_earnings
##  Min.   :   1.0    Min.   :1.000e+00                Min.   :     0         
##  1st Qu.:  27.0    1st Qu.:2.014e+07                1st Qu.:  2700         
##  Median :  65.5    Median :6.408e+07                Median : 13300         
##  Mean   : 745.7    Mean   :1.756e+08                Mean   : 36886         
##  3rd Qu.: 139.8    3rd Qu.:1.688e+08                3rd Qu.: 37900         
##  Max.   :7741.0    Max.   :6.589e+09                Max.   :850900         
##  NA's   :33        NA's   :56                                              
##  highest_monthly_earnings lowest_yearly_earnings highest_yearly_earnings
##  Min.   :       0         Min.   :       0       Min.   :        0      
##  1st Qu.:   43500         1st Qu.:   32650       1st Qu.:   521750      
##  Median :  212700         Median :  159500       Median :  2600000      
##  Mean   :  589808         Mean   :  442257       Mean   :  7081814      
##  3rd Qu.:  606800         3rd Qu.:  455100       3rd Qu.:  7300000      
##  Max.   :13600000         Max.   :10200000       Max.   :163400000      
##                                                                         
##  subscribers_for_last_30_days  created_year  created_month       created_date  
##  Min.   :      1              Min.   :1970   Length:995         Min.   : 1.00  
##  1st Qu.: 100000              1st Qu.:2009   Class :character   1st Qu.: 8.00  
##  Median : 200000              Median :2013   Mode  :character   Median :16.00  
##  Mean   : 349079              Mean   :2013                      Mean   :15.75  
##  3rd Qu.: 400000              3rd Qu.:2016                      3rd Qu.:23.00  
##  Max.   :8000000              Max.   :2022                      Max.   :31.00  
##  NA's   :337                  NA's   :5                         NA's   :5      
##  Gross.tertiary.education.enrollment....   Population        Unemployment.rate
##  Min.   :  7.60                          Min.   :2.025e+05   Min.   : 0.750   
##  1st Qu.: 36.30                          1st Qu.:8.336e+07   1st Qu.: 5.270   
##  Median : 68.00                          Median :3.282e+08   Median : 9.365   
##  Mean   : 63.63                          Mean   :4.304e+08   Mean   : 9.279   
##  3rd Qu.: 88.20                          3rd Qu.:3.282e+08   3rd Qu.:14.700   
##  Max.   :113.10                          Max.   :1.398e+09   Max.   :14.720   
##  NA's   :123                             NA's   :123         NA's   :123      
##  Urban_population       Latitude        Longitude      
##  Min.   :    35588   Min.   :-38.42   Min.   :-172.10  
##  1st Qu.: 55908316   1st Qu.: 20.59   1st Qu.: -95.71  
##  Median :270663028   Median : 37.09   Median : -51.93  
##  Mean   :224214982   Mean   : 26.63   Mean   : -14.13  
##  3rd Qu.:270663028   3rd Qu.: 37.09   3rd Qu.:  78.96  
##  Max.   :842933962   Max.   : 61.92   Max.   : 138.25  
##  NA's   :123         NA's   :123      NA's   :123

It can find that most numeric variables, like subscribers and lowest_monthly_earnings are highly right-skewed.

3 Data cleaning and transformation

3.1 Ckeck the columns names and uniform column name format

Use the names function to get the names of each column in data

names(youtube_data)

##  [1] "rank"                                   
##  [2] "Youtuber"                               
##  [3] "subscribers"                            
##  [4] "video.views"                            
##  [5] "category"                               
##  [6] "Title"                                  
##  [7] "uploads"                                
##  [8] "Country"                                
##  [9] "Abbreviation"                           
## [10] "channel_type"                           
## [11] "video_views_rank"                       
## [12] "country_rank"                           
## [13] "channel_type_rank"                      
## [14] "video_views_for_the_last_30_days"       
## [15] "lowest_monthly_earnings"                
## [16] "highest_monthly_earnings"               
## [17] "lowest_yearly_earnings"                 
## [18] "highest_yearly_earnings"                
## [19] "subscribers_for_last_30_days"           
## [20] "created_year"                           
## [21] "created_month"                          
## [22] "created_date"                           
## [23] "Gross.tertiary.education.enrollment...."
## [24] "Population"                             
## [25] "Unemployment.rate"                      
## [26] "Urban_population"                       
## [27] "Latitude"                               
## [28] "Longitude"

The columns names are not have the same format then use function to uniform the format.

Use the clean_names function in the janitor package to normalize the column names of the data frame

youtube_data <- youtube_data |> janitor::clean_names()
print(names(youtube_data))

##  [1] "rank"                                "youtuber"                           
##  [3] "subscribers"                         "video_views"                        
##  [5] "category"                            "title"                              
##  [7] "uploads"                             "country"                            
##  [9] "abbreviation"                        "channel_type"                       
## [11] "video_views_rank"                    "country_rank"                       
## [13] "channel_type_rank"                   "video_views_for_the_last_30_days"   
## [15] "lowest_monthly_earnings"             "highest_monthly_earnings"           
## [17] "lowest_yearly_earnings"              "highest_yearly_earnings"            
## [19] "subscribers_for_last_30_days"        "created_year"                       
## [21] "created_month"                       "created_date"                       
## [23] "gross_tertiary_education_enrollment" "population"                         
## [25] "unemployment_rate"                   "urban_population"                   
## [27] "latitude"                            "longitude"

3.2 Find invaild and missing values then tranform them

Find invalid value 0 in numerical variables, and find missing values “NaN” and “nan” in categorical variables, convert them to NA

youtube_data[youtube_data == 0 | youtube_data == "nan" | youtube_data == "NaN"] <- NA

3.3 Remove some invalid rows

According to the observation data, some invalid rows can be found. For example, the number of subscriptions is tens of thousands, but the number of views and uploaded videos is 0, and the corresponding income is also 0. This data should be collected for commercial protection reasons, so these data are somewhat distorted

# Find those colunms
variables_to_check <- c("video_views", "uploads", "lowest_monthly_earnings", 
                         "highest_monthly_earnings", "lowest_yearly_earnings", 
                         "highest_yearly_earnings")

# Find rows that meet both conditions
rows_with_all_na <- which(rowSums(is.na(youtube_data[, variables_to_check])) == length(variables_to_check))

# Delete these rows
deleted_rows <- youtube_data[rows_with_all_na, ]
deleted_values <- deleted_rows[, variables_to_check]
youtube_data_cleaned <- youtube_data[-rows_with_all_na, ]

# Print all deleted rows
print(deleted_values)

##     video_views uploads lowest_monthly_earnings highest_monthly_earnings
## 6            NA      NA                      NA                       NA
## 13           NA      NA                      NA                       NA
## 103          NA      NA                      NA                       NA
## 361          NA      NA                      NA                       NA
## 593          NA      NA                      NA                       NA
##     lowest_yearly_earnings highest_yearly_earnings
## 6                       NA                      NA
## 13                      NA                      NA
## 103                     NA                      NA
## 361                     NA                      NA
## 593                     NA                      NA

By analyzing the created_year column that needs to be used later, i found the existence of outlier

ggplot(youtube_data_cleaned, aes(x = created_year)) +
  geom_boxplot(outlier.colour="darkred",outlier.shape=18, outlier.size=8) +
  labs(title = "Box Plot of Created Year",
       x = "Year",
       y = "Created Year") +
  plot_theme + coord_flip()

It can be found that the outlier shows that it is 1970, and this year is an outlier, so this line of data has no reference significance. It should be deleted.

q1 <- quantile(youtube_data_cleaned$created_year, 0.25, na.rm = TRUE)
q3 <- quantile(youtube_data_cleaned$created_year, 0.75, na.rm = TRUE)
iqr <- q3 - q1
lower_bound <- q1 - 1.5 * iqr
upper_bound <- q3 + 1.5 * iqr

youtube_data_cleaned <- youtube_data_cleaned[youtube_data_cleaned$created_year >= lower_bound & youtube_data_cleaned$created_year <= upper_bound, ]

3.4 Process some columns

3.4.1 Handling NA values in channel type

Convert all NA values in the channel type column to the string “missing”.

youtube_data_cleaned$channel_type[is.na(youtube_data_cleaned$channel_type)] <- "missing"

3.4.2 Handling all NAs in numeric variables

For the purpose of maintaining data integrity, retaining sample size, retaining data structure, reducing bias, etc., convert NA in numerical variables into the median of the corresponding variable using function.

impute_na_with_median <- function(data) {
  for (col_name in names(data)) {
    col <- data[[col_name]]
    
    if (is.numeric(col)) {
      median_val <- median(col, na.rm = TRUE)
      col[is.na(col)] <- median_val
      data[[col_name]] <- col
    }
  }
  return(data)
}
youtube_data_cleaned <- impute_na_with_median(youtube_data_cleaned)

3.4.3 Delete columns

To make subsequent data analysis more convenient, delete some unnecessary columns.

youtube_data_transformed <- youtube_data_cleaned |>
  select(-category, -title, -abbreviation,
         -video_views_rank, -country_rank, -channel_type_rank, -created_date,
         -population, -unemployment_rate, -gross_tertiary_education_enrollment,
         -urban_population, -latitude, -longitude)

3.4.4 Data transformation

Transform all the categorical variables into factors using function

convert_to_factor <- function(data) {
  for (col_name in names(data)) {
    if (is.character(data[[col_name]])) {  
      data[[col_name]] <- as.factor(data[[col_name]])
    }
  }
  return(data)
}
youtube_data_transformed <- convert_to_factor(youtube_data_transformed)

4 Visualisation

4.1 Analyze channel type

Look at the distribution of YouTube channel

sorted_youtube_data <- youtube_data_transformed
channel_counts <- table(sorted_youtube_data$channel_type)
sorted_channel_types <- names(sort(channel_counts))

# Create a table of channel type counts
channel_counts <- table(sorted_youtube_data$channel_type)

# Convert channel_counts into a data frame for easier manipulation
channel_counts_df <- data.frame(channel_type = names(channel_counts), count = as.numeric(channel_counts))

# Sort the channel counts data frame by count
sorted_channel_counts_df <- channel_counts_df[order(channel_counts_df$count, decreasing = FALSE), ]


ggplot(sorted_channel_counts_df, aes(x = factor(channel_type, levels = sorted_channel_counts_df$channel_type), y = count)) +
  geom_bar(stat = "identity",fill = 'grey90',color = 'darkred',bins = 20) +
  geom_text(aes(label = count), vjust = -0.5,color = 'darkred') +
  labs(title = "Distribution of Channel Type",
       x = "Channel Type") + plot_theme + 
  theme(axis.text.x = element_text(angle = 45, hjust = 1,size = 10),
        axis.text.y = element_text(hjust = 0.5,size = 10),
        plot.title = element_text(size = 18))

It can be found that the proportion of music channels and entertainment channels is the largest, indicating that the YouTube platform focuses on relaxation and entertainment for people. Although the proportion of educational channels is not large, it is almost equal to that of comedy channels, which shows that some people use YouTube to learn knowledge.

4.2 Use shiny app to display scatter plot, density plot and map

world_map <- map_data("world")

Video_Views_by_Country <- youtube_data_transformed |>
  group_by(country) |>
  summarise(videoviews = sum(video_views))


map_data <- world_map |>
  left_join(Video_Views_by_Country, by = c("region" = "country"), copy = TRUE)


ui <- fluidPage(
  titlePanel("Shiny App"),
  
  verticalLayout(
   wellPanel(
      selectInput("plot_type", "Select Plot Type:",
                  choices = c("scatter_plot", "density_plot", "map")),
    
    
    # When select different plot types, different widgets will be showed
    conditionalPanel(
      condition = "input.plot_type == 'scatter_plot'",
      verticalLayout(
        selectInput("numeric_var", "Select Numeric Variable:",
                    choices = c("subscribers", "video_views")),
        plotOutput("scatter_plot", width = "800px", height = "600px")
      )
    ),
    conditionalPanel(
      condition = "input.plot_type == 'density_plot'",
      verticalLayout(
        radioButtons("earnings_var", "Select Earnings Variable:",
                     choices = c("monthly_earnings", "yearly_earnings")),
        plotOutput("density_plot", width = "800px", height = "400px")
      )
    ),
    conditionalPanel(
      condition = "input.plot_type == 'map'",
      verticalLayout(
        sliderInput(inputId = "slider_range", label = "Video Views Range",
                    min = 0, max = max(youtube_data_transformed$video_views),
                    value = c(0, max(youtube_data_transformed$video_views))),
        plotOutput("my_map", width = "800px", height = "600px")
      )
    )
  )
)
)


server <- function(input, output) {
  
  # Scatter plot
  output$scatter_plot <- renderPlot({
    x_var <- switch(input$numeric_var,
                    "subscribers" = "subscribers",
                    "video_views" = "video_views") 
     
    ggplot(youtube_data_transformed, aes_string(x = x_var, y = "channel_type", color = "channel_type")) +
      geom_point() +
      ggtitle(paste("Relationship between", input$numeric_var, "and Channel Type")) +
      ylab("Channel Type") +
      xlab(input$numeric_var) +
      theme_minimal() + plot_theme +
      theme(axis.title = element_text(size = 14),
        axis.text = element_text(size = 12),
        plot.title = element_text(size = 17,hjust = 0.5),
        legend.key.size = unit(1.2, "cm"),legend.text = element_text(size = 12),legend.title = element_text( size = 13))
  })
  
  # Density plot
  output$density_plot <- renderPlot({
    earnings_var <- input$earnings_var
    
    if (earnings_var == "monthly_earnings") {
      p <- ggplot(data = youtube_data_transformed) +
        geom_density(aes(x = log(lowest_monthly_earnings), fill = "Lowest Monthly Earnings"), alpha = 0.5) +
        geom_density(aes(x = log(highest_monthly_earnings), fill = "Highest Monthly Earnings"), alpha = 0.5) +
        scale_x_log10() +
        ggtitle("Monthly Earnings") +
        xlab("Earnings(log)") +
        ylab("Density") +
        labs(fill = "Variable") +
        plot_theme +
        theme(axis.title = element_text(size = 14),
        axis.text = element_text(size = 12),
        plot.title = element_text(size = 17,hjust = 0.5),
        legend.key.size = unit(1.2, "cm"),legend.text = element_text(size = 12),legend.title = element_text( size = 13))
      
      p + annotate("text", x = max(p$data$x), y = 0, label = "Lowest Monthly Earnings", vjust = 0, hjust = 1, color = "lightblue") +
          annotate("text", x = max(p$data$x), y = -0.02, label = "Highest Monthly Earnings", vjust = 0, hjust = 1, color = "darkred")
    } else if (earnings_var == "yearly_earnings") {
      p <- ggplot(data = youtube_data_transformed) +
        geom_density(aes(x = log(lowest_yearly_earnings), fill = "Lowest Yearly Earnings"), alpha = 0.5) +
        geom_density(aes(x = log(highest_yearly_earnings), fill = "Highest Yearly Earnings"), alpha = 0.5) +
        scale_x_log10() +
        ggtitle("Yearly Earnings") +
        xlab("Earnings(log)") +
        ylab("Density") +
        labs(fill = "Variable") + 
        plot_theme +
        theme(axis.title = element_text(size = 14),
        axis.text = element_text(size = 12),
        plot.title = element_text(size = 17,hjust = 0.5),
        legend.key.size = unit(1.2, "cm"),legend.text = element_text(size = 12),legend.title = element_text( size = 13))
      
      p + annotate("text", x = max(p$data$x), y = 0, label = "Lowest Yearly Earnings", vjust = 0, hjust = 1, color = "lightblue") +
          annotate("text", x = max(p$data$x), y = -0.02, label = "Highest Yearly Earnings", vjust = 0, hjust = 1, color = "darkred")
    }
  })
  
  # Map
  filtered_data <- reactive({
    subset(youtube_data_transformed,
           video_views >= input$slider_range[1] & video_views <= input$slider_range[2])
  })
  
  output$my_map <- renderPlot({
    filtered_map_data <- left_join(map_data, filtered_data(), by = c("region" = "country"))

    ggplot(data = filtered_map_data) +
      geom_polygon(aes(x = long, y = lat, group = group, fill = video_views / 1e9), color = "white") +
      scale_fill_gradient(name = "Video Views", low = "lightgrey", high = "darkred", labels = scales::comma) +
      coord_fixed(ratio = 1.3) +
      labs(title = "Video Views of Countries", fill = "Video Views") +
      scale_x_continuous(labels = scales::comma) +
      scale_y_continuous(labels = scales::comma) +
      theme_minimal() +
      theme(legend.position = "bottom", 
            plot.title = element_text(size = 20,hjust = 0.5,color = "darkred"),
            legend.key.size = unit(1.2, "cm"),
            legend.text = element_text(size = 12),
            legend.title = element_text( size = 13))
  })
}

shinyApp(ui, server)

Shiny applications not supported in static R Markdown documents

-Scatter plot

it can be seen that the distribution of subscriptions and views on different channel types is roughly the same. In terms of subscriptions, music, entertainment, games, and education channels have a relatively large number of subscribers, and music channels have the largest number of subscribers at 245 million. In terms of playback volume, it is mainly entertainment, music, and education channels that have a relatively large playback volume, and the largest playback volume is also created by the music channel.

-Density plot

Whether it is monthly earnings or annual income, the general trend of changes is similar. First the curve is longer for lower incomes and then rises steeply to a peak income. At the same time the peak of the lowest income is always lower than the peak of the highest income

-Map

By sliding the slider, you can see the video views in different intervals and the corresponding city distribution. It can be seen that the countries with high video views are mainly concentrated in India and Russia, because the blocks in these countries are darker

4.3 Analyze the distribution of channel types in different countries

Because the country data is too large and there are many countries with very small playback volume, i decided to extract the top 100 countries for analysis

new_youtube_data <- data.frame(
  Country = youtube_data_transformed$country,
  channel_type = youtube_data_transformed$channel_type
)

counting <- table(new_youtube_data$Country, new_youtube_data$channel_type)


counting_df <- as.data.frame(counting)
names(counting_df) <- c("Country", "Channel_type", "Count")


top_countries <- counting_df |>
  arrange(desc(Count)) |>
  head(100)

ggplot(data = top_countries, mapping = aes(x = Country, y = Channel_type)) +
  geom_count(aes(size = Count, fill = Count)) +
  scale_size_continuous(range = c(5, 20)) +
  scale_fill_gradient(low = "lightblue", high = "darkred") +
  ggtitle("Top 100 Countries vs. Channel Type") +
  labs(x = "Country", y = "Channel Type", fill = "Count") +
  plot_theme + coord_flip() +
  theme(panel.background = element_rect(fill = "gray"),  
        panel.grid.major = element_line(color = "white", linetype = 1),
        axis.text = element_text(size = 30,hjust = 1),
        axis.text.x = element_text(angle = 45),
        axis.title = element_text(size = 35),
        plot.title = element_text(size = 45),
        legend.key.size = unit(3, "cm"),
        legend.title = element_text(size = 25),
        legend.text = element_text(size = 25))

It can be found that most of these top 100 countries have a relatively large proportion of entertainment and music. At the same time, people in the two countries of united states and India watch the widest range of channels.

4.4 Analyze the subscription volume of different channel types in the past 30 days

custom_colors <- c("#1f78b4", "#33a02c", "#e31a1c", "#ff7f00", "#6a3d9a",
                   "#b15928", "#a6cee3", "#b2df8a", "#fb9a99", "#fdbf6f",
                   "#cab2d6", "#ffff99", "#8dd3c7", "#bebada", "#80b1d3")

ggplot(youtube_data_transformed, aes(x = channel_type, y = subscribers_for_last_30_days, fill = channel_type)) +
  geom_bar(stat = "identity") +
  scale_fill_manual(values = custom_colors) +  
  labs(title = "Subscribers for Last 30 Days by Channel Type",
       x = "Channel Type",
       y = "Numbers of Subscribers") +
  theme(axis.text = element_text(size = 25),
        axis.title = element_text(size = 35),
        legend.key.size = unit(3, "cm"),
        legend.text = element_text(size = 30),
        legend.title = element_text(size = 30),
        plot.title = element_text(size = 45,hjust = 0.5,color = "darkred"))+
  coord_polar() 

Similar to the distribution of the overall number of subscribers, the entertainment and music channels still have the largest number of subscriptions. The number of people following the people channel is second only to the number of people following the music channel.

4.5 Analyze the relationship between creat time and earning

4.5.1 First study the distribution of creation time

date_info <- data.frame(year = youtube_data_transformed$created_year,
                         month = youtube_data_transformed$created_month)
date_info <- na.omit(date_info)

date_info <- date_info[order(date_info$year, date_info$month), ]

ggplot(date_info, aes(x = factor(year), fill = factor(month))) +
  geom_bar() +
  scale_x_discrete(labels = unique(date_info$year), expand = c(0.05, 0)) +
  labs(title = "Distribution of creation time",
       x = "Year",
       y = "Count",
       fill = "Month") +
  plot_theme +
  theme(axis.text = element_text(size = 20), 
        axis.title = element_text(size = 23),
        plot.title = element_text(size = 30),
        legend.key.size = unit(1.5, "cm"),
        legend.text = element_text(size = 15),
        legend.title = element_text( size = 15))+
  scale_fill_brewer(palette = "Set3")

As can be seen from this bar chart, the number of channels created in the three years of 2006, 2011 and 2014 is far greater than in other years, and 2014 is the last year of large-scale growth. After 2014, the overall number of channel creations tended to decline, and even if there was an increase, it was only a small increase.

4.5.2 Analyze the relationship between creation time and video views

ggplot(data = youtube_data_transformed) +
geom_point(mapping = aes(y = video_views, x = created_year), color = "darkred",shape = 18,size = 4)+
geom_smooth(mapping = aes(y = video_views, x = created_year))+
  labs(title = "Created year vs. Video Views")+plot_theme +
  theme(plot.title = element_text(size = 20),
        axis.title = element_text(size = 15),
        axis.text = element_text(size = 13))

It can be seen that the number of views and created channels are directly in line with the development trend of channel creation. 2006 was not only the year when the number of created channels increased sharply, but also the year with the largest number of views. After that, the number of views remained at a certain level, accompanied by small-scale growth and slowdown.

4.5.3 Study the relationship between creation year and earning

count_data <- youtube_data_transformed |>
  group_by(created_year, channel_type) |>
  count()

p <- ggplot(count_data, aes(x = created_year, y = n, color = channel_type,text = paste("Channel Type: ", channel_type, "\nYear: ", created_year, "\nCount: ", n))) +
  geom_point() +  
  geom_line() +
  geom_point(size = 3, alpha = 0.6)+
  labs(title = " Created Year vs. Count by Channel Type",
       x = "Created Year",
       y = "Count",
       color = "Channel Type")+
  theme_minimal() +
  theme(axis.title = element_text(size = 12),
        axis.text = element_text(size = 11),
        plot.title = element_text(size = 17,hjust = 0.5,color = "darkred"),
        legend.key.size = unit(0.7, "cm"),
        legend.text = element_text(size = 10),
        legend.title = element_text( size = 13))

ggplotly(p, tooltip = "text")

We can move the cursor to view the specific channel type, year and corresponding number of videos uploaded by the channel at each point. It can be seen from the observation point map that among the channel types each year, the entertainment category is the type with the most videos uploaded by bloggers. At the same time, not all channel types are uploaded by bloggers every year.

4.5.4 Study the relationship between creation month and channel type

ggplot(youtube_data_transformed, aes(x = created_month, y = channel_type, color = channel_type)) +
  geom_point() +  
  geom_line(aes(group = channel_type)) + 
  labs(title = "Created Month vs. Channel Type",
       x = "Created Month",
       y = "Channel Type",
       color = "Channel Type")+
  theme_minimal() + 
  theme(axis.title = element_text(size = 20),
        axis.text = element_text(size = 16),
        plot.title = element_text(size = 27,hjust = 0.5,color = "darkred"),
        legend.key.size = unit(1, "cm"),
        legend.text = element_text(size = 16),
        legend.title = element_text( size = 19))

It can be seen that bloggers upload videos to every type of channel almost every month. Using only two channels, Nonprofit and Autos, there are almost no bloggers uploading videos for more than a few months.

5 Conclusion

The data set global youtube statisctic from Kaggle platform has 995 observations and 28 variables. There are missing values and invalid values in the data. It has been processed according to the corresponding analysis requirements and can be used for analysis.

In the analysis data, the analysis related to channel type, whether it is analyzing the type of channel type, or analyzing the relationship with the number of broadcasts or subscriptions, is the type of entertainment that is the most outstanding.

A clear trend can be seen in the income variable, that is, the income level of top bloggers is quite high, and the income of small bloggers is meager. Whether the amount of income is related to the time of creating accounts is worth further exploration.

Regarding the relationship between the number of broadcasts and the time of creating an account, it can be seen that there is a bonus period, but even if there is an increase after the bonus period, it will only increase slightly. It is reasonable to speculate that the earlier the account is created, the more likely it is to become a top blogger.

6 Reference

Holtz, Y. (n.d.). The R Graph Gallery – Help and inspiration for R charts. The R Graph Gallery. https://r-graph-gallery.com/

OpenAI. (2023). ChatGPT (August 3 Version) [Large language model]. https://chat.openai.com